comparative value priorities of chinese and new zealand

negative indicator such as no, not, never, etc. The item is expected to elicit disagreement
with the item from a respondent who has positive feelings about the concept, but the
psychodynamics of language seem to make this difficult or impossible. For example, if
a respondent feels comfortable using computers, he or she is expected to agree with the
statement “I find that using computers makes my work easier.”, and to disagree with the
statement “I find working with a computer to be stressful.” Reflecting on this item, we
begin to see the problem; we might expect spread of effect from stressful work to
influence response to the second item. Using a computer makes work easier but the
work is still stressful. The second item is not actually a reversed statement of the first.
The LBDQXII contains 20 items of 100 that are reverse-scored (21 for the Chinese
version). Use of reverse-scored items has proven to be problematic in survey research.
In an effort to reduce response set, or response bias, measurement some self-styled
experts recommended using negatively and positively worded items when measuring
the same construct. Nunnaly (1967) published an early work on response set bias (RSB,
bias being perhaps a misleading choice of words). RSB occurs when respondents do not
discriminate amongst survey items and respond to every question in the same manner
(e.g. circle all threes or fours on a five-point Likert scale). To reduce or eliminate RSB
by ensuring that the respondents were reading the questions in a thoughtful manner,
Nunnaly recommended negatively worded items should be included in survey
questionnaires. Nunnaly’s argument was that subjects who failed to recognize the
reversal of the items could be identified as engaging in some type of response bias and
could be removed from the sample, therefore providing a method by which researchers
could increase the accuracy of the data being analysed. The logic of reversing the
wording of particular items and then recoding them during scoring to be consistent with
the remaining items has a certain intuitive, though fallacious, appeal. All studies I
reviewed are not cited in this paper, however, comments of researchers on the low
correlations and factor loadings of reverse scored item almost universally indicate they
have a lower contribution to measurement of the concept of interest, sometimes to the
degree of lowering the scale reliability to an unacceptable level. Ibrahim (2001) believes
reverse-scored items in fact lead to tendencies such as acquiescence, malicious random
responding, and response set bias. Frequently, removing subjects who respond
differently to reverse-scored items would lead to unacceptably high losses of
participants.
314